From: Robert Myers on
Thanks to all who contributed their thoughts on high-bandwidth computing.

A wiki and mailing list are on their way for those who wish to pursue
the discussion.

I have previously posted a link here to this article

On the Effects of Memory Latency and Bandwidth on Supercomputer.
Application Performance. Richard Murphy. Sandia National Laboratories...
www.sandia.gov/~rcmurph/doc/latency.pdf

Google groups search is so broken that I can't find my previous post or
the discussion around it.

Robert.
From: Andy Glew "newsgroup at on
On 7/25/2010 11:49 PM, Brett Davis wrote:
> The one problem set I know of that really wants small random
> access is dictionary based AI, the last remaining approach
> to emulating the human mind, as all the other approaches have
> failed. (Think ELisa with a terra-byte database.)

Network routing (big routers, lots of packets).

People who are modelling large social networks, large networks of
actors. Inter-relationships.

From: Terje Mathisen "terje.mathisen at on
Brett Davis wrote:
> Intelligence would seem to be nothing more than a critical mass of
> data plus connections, allowed to self explore and learn. Quite a shock.

I believe intelligence (in humans) are very closely correlated with the
size of your associative memory, i.e. how large a problem you can grok
at once.
>
> No God or soul would seem to be needed, those are created afterwords,
> to explain that which cannot be explained. (This is a old idea, it
> comes up often in Bible studies.)

Just like creation myths, these ideas seemed to crop up in every society.

It is still extremely saddening to me to keep reading articles like the
one yesterday about a town in Louisiana where the school board was
unanimous in wanting "creative design" alongside or instead of evolution.

Terje
--
- <Terje.Mathisen at tmsw.no>
"almost all programming can be viewed as an exercise in caching"
From: Terje Mathisen "terje.mathisen at on
Benny Amorsen wrote:
> Andy Glew<"newsgroup at comp-arch.net"> writes:
>
>> Network routing (big routers, lots of packets).
>
> With IPv4 you can usually get away with routing on the top 24 bits + a
> bit of special handling of the few local routes on longer than 24-bit.
> That's a mere 16MB table if you can make do with possible 256
> "gateways".

Even going to a full /32 table is just 4GB, which is very cheap these
days. :-)

The alternative of multi-layer tables will save a lot of memory, but
double the number of lookups:

First-level table is /16, returning a 16-bit number which is either a
route/interface/gateway or a pointer to one of the secondary tables.

This way the 128K first-level table fits nicely in processor cache, and
handles almost all the long-distance routes, while the secondary tables
can be of adaptive size, i.e. the header contains the number of bits to
use for lookup.

The one problem with this approach is that it has variable latency, so
that makes it harder to use in a pure HW fast path. :-(

> You can simply replicate the whole routing table to local memory on all
> processors; updates are a challenge but it's usually ok if packets go to
> the wrong route for a few ms.

Right, the routing update could have arrived a few ms sooner or later
anyway.

Terje

--
- <Terje.Mathisen at tmsw.no>
"almost all programming can be viewed as an exercise in caching"
From: nmm1 on
In article <lre2i7-65g.ln1(a)ntp.tmsw.no>,
Terje Mathisen <"terje.mathisen at tmsw.no"> wrote:
>
>It is still extremely saddening to me to keep reading articles like the
>one yesterday about a town in Louisiana where the school board was
>unanimous in wanting "creative design" alongside or instead of evolution.

It's bloody terrifying! Those imbeciles seem to have read the first
and last books of the bible (in Authorized Version order), taken them
as gospel, and ignored everything in between, including the Gospels.
And they are coming very close to controlling most of the military
power of the planet, including almost all the nuclear weapons.


Regards,
Nick Maclaren.